Method and device for constructing multi-factor optimization knowledge base of high-energy laser system

Abstract

This application provides a method and apparatus for constructing a multi-factor optimization knowledge base for a high-energy laser system, belonging to the field of data processing technology. The method includes: acquiring a target dataset, which includes multiple parameters associated with the high-energy laser system, each parameter corresponding to multiple parameter samples; determining key parameters and their influence degrees from the multiple parameters based on the parameter samples, where the influence degree indicates the independent influence of each key parameter and the contribution of the interaction between key parameters to operational performance; determining multi-objective optimization objectives and engineering constraints; performing multi-objective optimization based on the key parameters, influence degrees, multi-objective optimization objectives, and engineering constraints to obtain the optimal parameter combination; and constructing a multi-factor optimization knowledge base based on the optimal parameter combination. This method can accurately locate the core influencing parameters of the high-energy laser system, improving system optimization efficiency and the accuracy of operational performance optimization.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, specifically relating to a method and apparatus for constructing a multi-factor optimization knowledge base for a high-energy laser system. Background Technology

[0002] In the application and optimization of high-energy laser systems, related technical solutions have long been limited by the constraints of multi-dimensional parameter analysis and the inefficiency of knowledge management, making it difficult to achieve precise improvement in system performance and effective reuse of optimization experience. Existing technologies often focus on singular analyses of scattered parameter samples, failing to accurately uncover the actual impact of different parameters on system performance or consider the coupling contribution of parameter interactions to operational effectiveness. This results in an inability to accurately identify key parameters that play a crucial role in operational performance, leading to a lack of direction in subsequent parameter optimization.

[0003] Meanwhile, in the parameter optimization stage, the lack of a scientific optimization strategy that incorporates the constraints of actual engineering applications makes it difficult to achieve a balance among multi-dimensional performance requirements. The optimized parameter combinations often fail to simultaneously consider engineering feasibility and optimal operational performance. Furthermore, existing technologies lack a structured system for managing the correlation between optimization data, parameter combinations, and operational effects of high-energy laser systems. Various optimization-related information is stored in a scattered manner, failing to form a quickly searchable and reusable knowledge system. Each time a new operational requirement arises, parameter analysis and optimization work must be carried out anew. This not only significantly increases the time and manpower costs of system optimization but also prevents past optimization experience from being effectively transformed into technical support for practical applications, severely hindering the improvement of high-energy laser system operational efficiency and optimization levels. Summary of the Invention

[0004] This application provides a method and apparatus for constructing a multi-factor optimization knowledge base for high-energy laser systems, which can accurately locate the core influencing parameters of high-energy laser systems, making multi-objective optimization more targeted and scientific, and improving the system optimization efficiency and the accuracy of operational performance optimization.

[0005] This application provides a method for constructing a multi-factor optimization knowledge base for high-energy laser systems, including: Obtain the target dataset, which includes multiple parameters associated with the high-energy laser system, with each parameter corresponding to multiple parameter samples; Based on the multiple parameter samples, key parameters and their parameter influence are determined from the multiple parameters. The parameter influence indicates the degree of contribution of the independent influence of each key parameter and the interaction between key parameters to the job performance. Determine the multi-objective optimization objectives and engineering constraints; Based on the key parameters, the parameter influence, the multi-objective optimization objective, and the engineering constraints, multi-objective optimization is performed to obtain the optimal combination of objective parameters. A multi-factor optimization knowledge base is constructed based on the optimal combination of parameters for the target.

[0006] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the step of determining key parameters and their parameter influence from the multiple parameter samples includes: obtaining the main effect value of each parameter from the multiple parameter samples corresponding to each parameter, wherein the main effect value is used to indicate the independent influence of each parameter; determining reference parameters based on the main effect values; determining the interaction variance of each reference parameter pair from the multiple parameter samples corresponding to the reference parameters, wherein the interaction variance is used to indicate the contribution of the interaction between the reference parameter pairs to the operational performance, wherein the reference parameter pair includes any two reference parameters; and determining key parameters from the reference parameters based on the interaction variance, wherein the parameter influence of the key parameters includes the main effect value and the interaction variance.

[0007] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the step of obtaining the main effect value of each parameter based on multiple parameter samples corresponding to each parameter includes: performing the following operations for each parameter to obtain the main effect value of each parameter: dividing the value of the current parameter into a high value level and a low value level; obtaining the number of first parameter samples with values ​​at the high value level and the number of second parameter samples with values ​​at the low value level among the multiple current parameter samples; obtaining the mean of the first operational effect coefficient at the high value level and the mean of the second operational effect coefficient at the corresponding low value level for the current parameter, wherein the operational effect coefficient is used to indicate the effect of the laser on the target; and determining the main effect value of the current parameter based on the mean of the first operational effect coefficient, the mean of the second operational effect coefficient, the number of first parameter samples, and the number of second parameter samples.

[0008] According to the multi-factor optimization knowledge base construction method for high-energy laser systems provided in this application, the step of determining the interaction variance of each reference parameter pair based on multiple parameter samples corresponding to the reference parameter includes: obtaining the operational performance parameters of each reference parameter sample corresponding to the reference parameter; determining the first operational performance parameter mean of each reference parameter based on the operational performance parameters of each reference parameter sample; obtaining the number of value levels of each reference parameter and the parameter sample corresponding to each value level; obtaining the second operational performance parameter mean of each reference parameter at each value level based on the parameter sample corresponding to each value level; and determining the interaction variance of each reference parameter based on the first operational performance parameter mean and the second operational performance parameter mean. The single-parameter variance of each reference parameter is determined by the value and the number of value levels; the level combinations of each reference parameter pair are obtained based on the value levels of each reference parameter, as well as the number of level combinations of the reference parameter pairs; the mean of the third job performance parameter of each level combination and the mean of the fourth job performance parameter of each reference parameter pair are obtained based on the job performance parameters of each reference parameter sample; the parameter variance is determined based on the mean of the third job performance parameter, the mean of the fourth job performance parameter, and the number of level combinations; the interaction variance of each reference parameter pair is determined based on the single-parameter variance of the reference parameters included in each reference parameter pair and the parameter variance of the reference parameter pair.

[0009] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the step of determining key parameters from the reference parameters based on the interaction variance includes: obtaining the operational performance parameters of each reference parameter sample corresponding to the reference parameters and a first number of reference parameter samples; determining the mean of a fourth operational performance parameter of all reference parameter samples based on the operational performance parameters of each reference parameter sample; determining the total variance of the reference parameters based on the first number, the operational performance parameters of each reference parameter sample, and the mean of the fourth operational performance parameter; determining the individual parameter variance of each reference parameter; obtaining the error variance of all reference parameters; determining the total interaction variance of the reference parameters based on the total variance of the reference parameters, the individual parameter variances, and the error variances; and determining key parameters from the reference parameters based on the interaction variance, the total variance of the reference parameters, and the total interaction variance.

[0010] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the multi-objective optimization based on the key parameters, the parameter influence degree, the multi-objective optimization objective, and the engineering constraints includes: constructing a regression equation; obtaining the actual operational performance parameters of each key parameter sample corresponding to the key parameters; calculating the predicted operational performance parameters of each key parameter sample based on the regression equation; optimizing the regression equation based on the predicted operational performance parameters and the actual operational performance parameters to obtain a target regression equation, which is used to predict the operational performance parameters of the parameter samples; obtaining a single-objective fitness function based on the historical best operational performance parameters, the maximum allowable energy consumption, and the longest allowable operational time; and performing multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the target regression equation to obtain the optimal combination of target parameters.

[0011] According to the multi-factor optimization knowledge base construction method for high-energy laser systems provided in this application, the step of performing multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the objective regression equation to obtain the optimal parameter combination includes: obtaining multiple key parameter combinations based on the key parameters; obtaining the parameter contribution rate of each key parameter; encoding each key parameter combination into a chromosome based on the parameter contribution rate to obtain an initial population; and substituting the key parameter combination corresponding to each chromosome in the current population into the objective regression equation to obtain the reference operational performance parameters output by the objective regression equation. In the initial iteration, the current population is the initial population, and the reference operational performance parameters include operational effect coefficients. The process involves considering factors such as the number of targets, operation time, or energy consumption, with the operation effect coefficient indicating the laser's effect on the target. A reference fitness value is determined based on the single-target fitness function according to the reference operation performance parameters. The next generation population is determined based on the reference fitness value. The next generation population is updated using a single-point crossover strategy and a bit mutation strategy. The next generation population is identified as the current population. The above steps are iteratively executed until a preset convergence condition is met. This preset convergence condition includes that the change in the reference fitness value of consecutive preset generations of populations in terms of the operation effect coefficient, operation time, and / or energy consumption is less than a second preset value. The optimal combination of target parameters is determined based on the chromosomes in the last generation population and the engineering constraints.

[0012] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the method further includes: after iteratively executing the above steps for the Nth time, if the preset convergence condition is not met, increasing the current population size and increasing the crossover probability of the single-point crossover strategy, and continuing to iterate for M times until the preset convergence condition is met, where N and M are positive numbers; if the preset convergence condition is not met after continuing to iterate for M times, using a simulated annealing algorithm to optimize and iteratively execute the above steps again based on the optimal parameter combination obtained in the Mth iteration.

[0013] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the step of obtaining the target dataset includes: constructing a digital twin of the high-energy laser system; obtaining experimental data and target simulation data of the digital twin; and generating the target dataset based on the experimental data and the target simulation data.

[0014] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the step of constructing the multi-factor optimization knowledge base based on the target optimal parameter combination includes: obtaining scene labels for the target optimal parameter combination; inputting the target optimal parameter combination into the digital twin based on the scene labels for multiple simulation verifications to obtain target performance data of the target optimal parameter combination; and constructing the multi-factor optimization knowledge base based on the target optimal parameter combination, the scene labels, and the target performance data.

[0015] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, after constructing the multi-factor optimization knowledge base based on the target optimal parameter combination, the method further includes: obtaining the current optimal parameter combination corresponding to the scene label; inputting the current optimal parameter combination into the digital twin based on the scene label for multiple simulation verifications to obtain the current performance data of the current optimal parameter combination; and updating the multi-factor optimization knowledge base based on the current optimal parameter combination, the scene label, and the current performance data.

[0016] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the construction of a digital twin of the high-energy laser system includes: acquiring geometric modeling data, a physical attribute parameter table, and a behavioral logic requirement file; constructing a geometric twin model based on the geometric modeling data; configuring physical twin characteristics based on the physical attribute parameter table; performing twin control based on the behavioral logic requirement file to obtain twin control results; and constructing the digital twin based on the geometric twin model, the physical twin characteristics, and the twin control results.

[0017] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the geometric modeling data includes three-dimensional data of the laser operation system, three-dimensional data of the target object, and component design drawings. The step of constructing a geometric twin model based on the geometric modeling data includes: generating an initial model of the laser operation system based on the three-dimensional data of the laser operation system; simplifying the initial model of the laser operation system into a mesh to obtain a simplified model of the laser operation system; calibrating and simulating the simplified model of the laser operation system based on the component design drawings to obtain a geometric twin model of the laser operation system; generating an initial model of the target object based on the three-dimensional data of the target object; configuring the size parameters of the target object; generating a geometric twin model of the target object based on the size parameters and the initial model of the target object; and constructing a geometric twin model based on the geometric twin model of the laser operation system and the geometric twin model of the target object.

[0018] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, the experimental data includes physical experimental data and empirical data. The acquisition of experimental data and the target simulation data of the digital twin includes: acquiring real-time twin simulation data, output signals of experimental sensors, and experimental logs; parsing the JSON fields of the real-time twin simulation data to obtain the target simulation data; converting the output signals into target digital signals to obtain physical experimental data; and obtaining key-value pairs from the experimental logs to obtain the empirical data.

[0019] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, before generating the target dataset based on the experimental data and the target simulation data, the method further includes: determining whether there is a conflict between the empirical data and the target simulation data or the entity experimental data; if there is a conflict, determining the credibility score of the target simulation data or the entity experimental data; and deleting the empirical data if the credibility score is higher than a preset score.

[0020] According to the method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application, a target dataset is generated based on the experimental data and the target simulation data, including: fusing the experimental data and the target simulation data to obtain fused data; obtaining scene labels; performing deduplication processing on the fused data based on the scene labels to obtain reference fused data; and preprocessing the reference fused data to obtain the target dataset, wherein the preprocessing includes outlier removal and missing value imputation.

[0021] This application also provides a device for constructing a multi-factor optimization knowledge base for a high-energy laser system, including: An acquisition unit is used to acquire a target dataset, which includes multiple parameters associated with a high-energy laser system, and each parameter corresponds to multiple parameter samples; The first determining unit is configured to determine key parameters and the parameter influence degree of the key parameters from the multiple parameter samples, wherein the parameter influence degree indicates the degree of contribution of the independent influence degree of each key parameter and the interaction between key parameters to the job performance. The second determining unit is used to determine the multi-objective optimization objective and engineering constraints. A multi-objective optimization unit is used to perform multi-objective optimization based on the key parameters, the parameter influence degree, the multi-objective optimization objectives and the engineering constraints, to obtain the optimal combination of objective parameters. The construction unit is used to construct a multi-factor optimization knowledge base based on the optimal combination of the target parameters.

[0022] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multi-factor optimization knowledge base construction method for any of the high-energy laser systems described above.

[0023] This application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the multi-factor optimization knowledge base construction method for high-energy laser systems as described above.

[0024] This application also provides a computer program product, including a computer program that, when executed by a processor, implements a multi-factor optimization knowledge base construction method for any of the high-energy laser systems described above.

[0025] According to the method and apparatus for constructing a multi-factor optimization knowledge base for high-energy laser systems provided in this application, the following steps are taken: First, a target dataset covering multiple related parameters and corresponding samples of the high-energy laser system is obtained. Then, based on the parameter samples, key parameters and parameter influence degrees including independent influence degree and interaction contribution degree are accurately determined. Next, the multi-objective optimization objectives and engineering constraints are clarified. Subsequently, multi-objective optimization is carried out by combining key parameters, parameter influence degrees and objective constraints to obtain the optimal parameter combination. Finally, a multi-factor optimization knowledge base is constructed based on this. This method can accurately locate the core influencing parameters of the high-energy laser system, making multi-objective optimization more targeted and scientific, ensuring that the optimization results meet the actual engineering needs, and realizing the structured accumulation and reuse of high-energy laser system optimization knowledge, which greatly improves the system optimization efficiency and the accuracy of operational performance optimization. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is one of the flowcharts illustrating a method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application.

[0028] Figure 2 This is the second flowchart of a method for constructing a multi-factor optimization knowledge base for a high-energy laser system provided in this application.

[0029] Figure 3 This is a block diagram of the functional units of a multi-factor optimization knowledge base construction device for a high-energy laser system provided in this application.

[0030] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0033] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0034] Existing technologies in the field of high-energy laser system application and optimization are limited by the singular and fragmented analysis of multi-dimensional parameters and the lack of scientific multi-objective optimization strategies that combine engineering constraints. As a result, they cannot accurately identify the core key parameters of system operation performance, nor can they obtain parameter combinations that take into account both engineering feasibility and optimal performance. This significantly increases optimization costs and severely restricts the improvement of system operation efficiency and optimization level.

[0035] To address the aforementioned issues, this application provides a method and apparatus for constructing a multi-factor optimization knowledge base for high-energy laser systems. The following detailed explanation of this application is provided in conjunction with the accompanying drawings.

[0036] Please see Figure 1 , Figure 1 This is one of the flowcharts illustrating a method for constructing a multi-factor optimization knowledge base for a high-energy laser system, as provided in this application. The method includes the following steps.

[0037] S101, Obtain the target dataset.

[0038] The target dataset includes multiple parameters associated with the high-energy laser system, with each parameter corresponding to multiple parameter samples. The parameters in the dataset comprehensively cover all relevant dimensions of the high-energy laser operation process, specifically including laser system parameters, target object parameters, environmental parameters, and operational performance index parameters. Laser system parameters may include output power, pulse width, and beam quality factor; target object parameters may include material thermal conductivity, melting point, specific heat capacity, surface roughness, movement speed, and attitude change characteristics; environmental parameters may include atmospheric visibility, turbulence intensity, and rain / fog particle density; and operational performance index parameters may include operational effectiveness coefficients, energy consumption, and operation duration.

[0039] S102, determine key parameters and the parameter influence degree of the key parameters from the multiple parameter samples.

[0040] The parameter influence degree indicates the degree of contribution of the independent influence of each key parameter and the interaction between key parameters to the job performance.

[0041] S103, Determine the multi-objective optimization objective and engineering constraints.

[0042] The multi-objective optimization objectives can be set in conjunction with the core requirements of high-energy laser operations. These objectives typically revolve around dimensions such as operational effectiveness and cost. For example, maximizing the operational effectiveness coefficient can be the core objective, combined with objectives such as minimizing energy consumption and minimizing operational time to form a multi-objective optimization system. Different weight coefficients can be assigned to different optimization objectives, with the weight coefficients summing to 1. The weight allocation can be flexibly adjusted according to the importance of each objective in the actual operational scenario, making the multi-objective optimization objectives more closely aligned with actual operational needs.

[0043] Engineering constraints are designed to limit the scope of parameter optimization and avoid parameter combinations that exceed system capabilities and operational requirements. These constraints can encompass multiple aspects, including hardware constraints and performance constraints. Hardware constraints determine the upper and lower limits of each parameter based on the actual hardware configuration of the high-energy laser system, clarifying the adjustable ranges of various parameters such as laser output power, stage movement speed, and target size adjustment range. This ensures that parameter values ​​do not exceed the actual support capabilities of the system hardware during the optimization process. Performance constraints set upper limits for energy consumption and operation time during the operation process, preventing excessive energy and time consumption due to an overemphasis on a single operational effect. Furthermore, other performance constraints can be supplemented based on the actual operational scenario.

[0044] S104. Based on the key parameters, the parameter influence, the multi-objective optimization objective, and the engineering constraints, perform multi-objective optimization to obtain the optimal parameter combination.

[0045] The optimal parameter combination indicates a high-energy laser system parameter configuration scheme that achieves comprehensive optimization of multiple objectives while meeting various engineering constraints. This combination represents the core parameter adaptation results for a specific operational scenario and can directly guide the setting of actual high-energy laser operation parameters. The optimal parameter combination can include specific values ​​of various key parameters identified in the early stages, specifically covering the optimal values ​​of key parameters related to the laser system, target object, and environment adaptation, such as the optimal values ​​for laser output power and pulse width, the adapted values ​​for target movement speed, and the values ​​of key environmental adaptation parameters determined by atmospheric visibility and turbulence intensity.

[0046] S105, Construct a multi-factor optimization knowledge base based on the optimal combination of the target parameters.

[0047] The multi-factor optimization knowledge base is used to achieve the association, storage, rapid querying, and efficient reuse of high-energy laser operation scenarios, key parameters, optimal parameter combinations, and operational effects. It can provide technicians with direct references for optimal parameter configurations for different high-energy laser operation scenarios, while also providing historical knowledge support for parameter optimization in new scenarios. It includes scenario tag data related to the high-energy laser system, the optimal parameter combinations for each scenario, and operational performance parameters verified by simulation. Scenario tag data can cover various scenario feature information such as target material, environmental conditions, and operational requirements, while operational performance parameters can include quantitatively verified indicators such as operational effect coefficients, energy consumption, and operational duration.

[0048] As can be seen, this solution first acquires a target dataset covering multiple related parameters of a high-energy laser system and their corresponding samples. Then, based on the parameter samples, it accurately determines key parameters and parameter influence degrees, including independent influence degrees and interaction contribution degrees. Next, it clarifies the multi-objective optimization objectives and engineering constraints. Subsequently, it combines key parameters, parameter influence degrees, and objective constraints to conduct multi-objective optimization to obtain the optimal parameter combination. Finally, it constructs a multi-factor optimization knowledge base, which can accurately locate the core influencing parameters of the high-energy laser system, making multi-objective optimization more targeted and scientific, ensuring that the optimization results meet the actual engineering needs, and realizing the structured accumulation and reuse of high-energy laser system optimization knowledge, significantly improving system optimization efficiency and the accuracy of operational performance optimization.

[0049] In one possible embodiment, acquiring the target dataset includes: constructing a digital twin of a high-energy laser system; acquiring experimental data and target simulation data of the digital twin; and generating the target dataset based on the experimental data and the target simulation data.

[0050] The target simulation data can include the operation effect coefficient, multiple independent simulation raw values ​​of energy consumption, and basic parameters of the simulation process such as laser system, target, and environment. The operation effect coefficient can be calculated in real time by the ray detection algorithm of the digital twin, and generated according to rules after being compared with the target material threshold. The energy consumption can be calculated by the twin in combination with the laser core parameters, operation time and atmospheric energy attenuation model.

[0051] The job performance coefficient is a quantitative evaluation result of a single simulation operation performed by the laser on the target. It ranges from 0 to 1, with a value closer to 1 indicating a better performance. It can be calculated and generated by a custom ray detection algorithm developed in Unity3D digital twin: During the simulation, the ray detection algorithm captures the three-dimensional coordinates of the laser's point of action in real time at a frequency of 100Hz and accurately calculates the energy density of the point of action. The real-time calculated energy density is compared with the target material's preset energy density threshold, which is determined by the target material's physical parameters such as melting point and thermal conductivity, and is pre-configured in the physical property parameter table. Based on the comparison results and according to the preset quantitative judgment rules, the original value of the job performance coefficient for a single simulation is automatically output.

[0052] In one possible embodiment, constructing a digital twin of a high-energy laser system includes: acquiring geometric modeling data, a physical attribute parameter table, and a behavioral logic requirement file; constructing a geometric twin model based on the geometric modeling data; configuring physical twin characteristics based on the physical attribute parameter table; performing twin control based on the behavioral logic requirement file to obtain twin control results; and constructing the digital twin based on the geometric twin model, the physical twin characteristics, and the twin control results.

[0053] The geometric modeling data may include 3D data of the laser operation system, 3D data of the target object, and design drawings of key components. The 3D data of the laser operation system can be obtained from the 3D point cloud data of the laser emitting unit and optical transmission unit acquired by a Faro laser scanner. The 3D data of the target object may include a 3D model file of the target entity generated by 3D design software (SolidWorks) and a corresponding Excel-formatted dimensional parameter configuration table. The physical property parameter table may cover laser parameter tables, material property tables, and environmental parameter tables. The laser parameter table includes output power range, pulse width range, and beam quality factor baseline values. The material property table includes parameters such as thermal conductivity, melting point, specific heat capacity, and surface roughness of the target material. The environmental parameter table includes atmospheric visibility baseline values ​​and turbulence intensity (atmospheric refractive index structure constant, Cn). 2 The requirements document includes the laser emission timing rules, platform movement speed range, and control mode. Target behavior requirements include the target movement mode, attitude change type, and triggering conditions. Interactive behavior requirements include the laser application point calculation frequency, energy density threshold, and operational effect coefficient determination rules.

[0054] When configuring physical twin characteristics, a custom shader can be created for the laser properties to implement a beam energy attenuation model. The power attenuation coefficient is then calculated based on atmospheric visibility, ensuring that laser energy transmission conforms to actual atmospheric laws. The power attenuation coefficient k = 0.02 × (5 / (V+1)) + 0.005, where V is atmospheric visibility, with an effective range of [0.5km, 20km]. In this case, k ∈ [0.0075~0.024] / km, ensuring that energy attenuation conforms to atmospheric laser transmission laws.

[0055] Simultaneously, the target material parameter table is imported into the Unity engine and bound to the corresponding target model through the Physical Material component to achieve simulation calculations of physical properties such as thermal conductivity and melting point. Atmospheric turbulence fields can also be generated using the Perlin noise algorithm, combined with a particle system to simulate rain and fog environments. By precisely configuring rain and fog particle density through the correlation between particle density and atmospheric visibility, the physical twin characteristics of the laser, target, and environment are configured in all dimensions, ensuring a high degree of matching between the physical properties of the digital twin and the actual scene. Turbulence intensity Cn 2 The correlation between Cn and the noise frequency f is Cn 2 =10 -13 ×(f+10), noise frequency f∈[1Hz,100Hz], corresponding to Cn 2 ∈[1.1×10 -12 1.01×10 -11 ]m -2 / 3 Suitable for medium to strong turbulence scenarios; for high turbulence scenarios, f can be extended to 500Hz, Cn 2 The upper limit has been adjusted to 5.01× 10-11 m -2 / 3 This formula is based on the fitting of measured near-surface atmospheric turbulence data and is applicable to working environments at altitudes ≤1000m. It simulates rain and fog environments using a particle system, and the negative correlation between particle density ρ and visibility V is given by ρ=max(10,-50×V+500), unit: particles / m. 3 For V∈[0.5km,9.8km], when V≥9.8km, ρ is fixed at 10 units / m. 3 Simulate clean atmospheric background particles to ensure non-negative density and conformity to the physical scenario.

[0056] By performing twin control and obtaining twin control results, and implementing behavioral twins, precise control of laser emission timing and stage movement can be achieved through scripting. It supports flexible adjustment of pulse width and stage movement speed, as well as switching between multiple control modes. Simultaneously, based on the Animator component, an animation controller for target movement and attitude changes can be created. It supports modifying the target movement speed and attitude change frequency through parameter configuration files. It can also develop ray detection algorithms and calculate the three-dimensional coordinates and energy density of the laser action point in real time at a preset frequency. Based on the comparison between energy density and target material threshold, the operation effect coefficient is output. This achieves twin control of the laser system, target object, and their interaction behavior, resulting in twin control results that fit the actual operation logic.

[0057] The laser emission timing control script is the core program for managing the timing and parameters of the entire high-energy laser emission process. Its core functions include: independently achieving precise digital control of laser emission, supporting programmed control of pulse width (adjustable in 0.1-10ms steps), emission frequency, duration, and instant start / stop; and automatically triggering laser emission upon receiving external commands, strictly matching the timing requirements of the simulation scenario.

[0058] The Target Behavior Script is a custom C# control script developed using the Unity3D engine and the Animator animation controller. It is a dedicated program for simulating the dynamic movement and posture changes of a target. It controls all dynamic behaviors of the target according to preset rules, including uniform / variable / curved movement modes and posture changes such as flipping / rotating / translating, while also binding trigger conditions. It supports real-time modification of movement speed and posture change frequency via configuration files, completely replicating the dynamic characteristics of a target in a real-world scene.

[0059] After completing the geometric twin model construction, physical twin characteristic configuration, and twin control, the three can be integrated to give the digital twin high-precision geometric shape, realistic physical characteristics, and behavior that conforms to the operation logic. Ultimately, a digital twin that can accurately map the actual operation state of a high-energy laser system can be constructed.

[0060] As can be seen, this solution first acquires basic data related to geometry, physics, and behavior, then sequentially completes the construction of a geometric twin model, the configuration of physical twin characteristics, and the implementation of twin control. Finally, it integrates the results of these three aspects to construct a digital twin of the high-energy laser system. This enables high-precision mapping of the system, environment, and target in terms of geometry, physics, and behavior, allowing the digital twin to accurately reproduce the actual operating state of the high-energy laser system. This provides a highly reliable virtual experimental scenario for subsequent parameter analysis and optimization verification, ensuring the accuracy and reliability of subsequent related work.

[0061] In one possible embodiment, the geometric modeling data includes 3D data of the laser operation system, 3D data of the target object, and component design drawings. The step of constructing a geometric twin model based on the geometric modeling data includes: generating an initial model of the laser operation system based on the 3D data of the laser operation system; simplifying the initial model of the laser operation system into a mesh to obtain a simplified model of the laser operation system; calibrating and simulating the simplified model of the laser operation system based on the component design drawings to obtain a geometric twin model of the laser operation system; generating an initial model of the target object based on the 3D data of the target object; configuring the size parameters of the target object; generating a geometric twin model of the target object based on the size parameters and the initial model of the target object; and constructing a geometric twin model based on the geometric twin model of the laser operation system and the geometric twin model of the target object.

[0062] Laser system modeling is based on the actual hardware of a high-energy laser operation system, and high-precision geometric twin modeling is carried out based on 3D point cloud data and design drawings of key components. Specifically, it may include the following steps: Import 3D point cloud data of the laser emitting unit and optical transmission unit collected by the Faro laser scanner, as well as design drawings of core components such as the laser resonator and focusing lens; import the point cloud data into the Unity3D engine, and simplify the mesh using the MeshLab tool, reducing the number of mesh faces by 60% while fully preserving the core structural features of the laser system; perform texture mapping on the simplified mesh model to restore the actual geometric appearance and structural details of the laser system; based on the design drawings of key components, perform geometric calibration on core parts such as the laser emission port and optical transmission path, and correct deviations through multiple simulations to ultimately ensure that the geometric error of the laser system twin model is ≤0.1mm, meeting the accuracy requirements of high-energy laser operation simulation.

[0063] Target object modeling is based on parametric 3D models and configurable dimension controls to construct geometric twin models of target objects adapted to multiple scenarios. It imports the target entity 3D model file generated by SolidWorks modeling, along with a corresponding Excel-formatted dimension parameterization configuration table containing adjustable fields such as length, width, and height. Model import and control configuration: The 3D model is imported into the Unity3D engine, and dimension parameterization controls are configured for the model through the Unity Inspector panel, enabling visual adjustment of the target's geometric dimensions. Users can input specific values ​​in the configuration table to customize adjustments to core dimensions such as length, width, and height, with an adjustment range of ±30%, adapting to target operation scenarios with different geometric dimensions. It ensures that the structural relevance and geometric integrity of the target model are not affected after parameterization adjustments, meeting the geometric simulation requirements for laser action point positioning and quantitative analysis of operational effects.

[0064] As can be seen, this solution obtains a high-precision geometric twin model of the laser operation system by modeling, mesh simplification, drawing calibration, and simulation verification of the 3D data of the laser operation system. At the same time, it generates a geometric twin model of the target object by modeling the 3D data of the target object and configuring adjustable size parameters. The two are then merged to construct a geometric twin model. This approach can optimize the model structure while retaining core features. Furthermore, the accuracy and flexibility of geometric modeling are improved through calibration verification and parameter configuration. This allows the constructed geometric twin model to accurately map the actual geometric shape of the laser operation system and the target object, ensuring the credibility of the geometric simulation.

[0065] In one possible embodiment, the experimental data includes physical experimental data and empirical data. The acquisition of experimental data and the target simulation data of the digital twin includes: acquiring real-time data of the twin simulation, output signals of experimental sensors, and experimental logs; parsing the JSON fields of the real-time twin simulation data to obtain the target simulation data; converting the output signals into target digital signals to obtain physical experimental data; and obtaining key-value pairs based on the experimental logs to obtain the empirical data.

[0066] The target simulation data of the digital twin can be a JSON-formatted structured data stream collected and output in real time during high-energy laser virtual simulation operations based on the Unity3D digital twin building module. It represents a dynamic record of all-dimensional operational parameters in the virtual experimental scenario and serves as the core virtual data source for multi-source data fusion, providing highly realistic virtual data support for subsequent data analysis and parameter optimization. The physical experiment data can include sensor-collected data and experimental record data. Sensor-collected data includes raw data collected in real time from core physical quantities during physical experiments conducted by the high-energy laser system using various professional industrial sensors and high-speed imaging equipment. This includes laser output power, target state data collected by high-speed cameras, and atmospheric visibility and turbulence intensity data collected by environmental sensors. All data includes a timestamp, enabling time-series alignment of data collected from multiple devices. Experimental record data can include metadata such as scene tags, personnel, and equipment numbers from the physical experiment.

[0067] Experience data can include experiential knowledge data and a scenario tag-parameter mapping table. Experience knowledge data can include engineer experience logs, containing descriptions of historical work scenarios, recommended parameter combinations, and feedback on work results. The scenario tag-parameter mapping table is a manually compiled mapping of typical scenarios to commonly used parameter combinations, including scenario ID, scenario description, parameter name, and recommended values.

[0068] When acquiring real-time twin simulation data, the JSON fields of the real-time twin simulation data can be parsed to obtain the target twin simulation data; the output signals of the experimental sensors can be read and converted into target digital signals. For physical experiment data acquisition, sensor output signals can be read using a sensor data acquisition card, converted into digital signals, and stored in .csv format, linked to the experimental record data via timestamps. For experiential knowledge data acquisition, a BERT model can be used to extract core information such as scene keywords, parameter combinations, and effect evaluations from the logs, generating standardized key-value pairs.

[0069] As can be seen, this solution obtains target simulation data, entity experimental data, and empirical data by acquiring real-time data from twin simulations, experimental sensor output signals, and experimental logs, parsing JSON fields in the simulation data, converting sensor signals into digital signals, and extracting key-value pairs from the experimental logs. This enables standardized acquisition and unified format processing of multi-source experimental and simulation data, accurately extracts the core effective information of various data types, and ensures the consistency and standardization of the acquired experimental data and target simulation data, laying a high-quality data foundation for subsequent data fusion processing.

[0070] In one possible embodiment, before generating the target dataset based on the experimental data and the target simulation data, the method further includes: determining whether there is a conflict between the empirical data and the target simulation data or the entity experimental data; if there is a conflict, determining the credibility score of the target simulation data or the entity experimental data; and deleting the empirical data if the credibility score is higher than a preset score.

[0071] The reliability score calculation comprehensively considers multiple dimensions, including the accuracy of data acquisition equipment, the stability of the acquisition environment, the consistency of sample repeatability, and the pass rate of simulation model verification. For physical experimental data, the score is based on sensor calibration accuracy, the degree of interference in the experimental environment, and the coefficient of variation from repeated experiments. For target simulation data, the score is based on the model verification accuracy of the digital twin, the matching degree between simulation parameters and actual working conditions, and the consistency between historical simulation results and measured results. Different weighting coefficients can be set for different influencing factors, and the final reliability score is obtained through weighted summation, ensuring that the score objectively reflects the true reliability of the data. When empirical knowledge conflicts with experimental data, simulation experimental data with a reliability score ≥ 0.8 is used as the standard, eliminating conflicting empirical data.

[0072] As can be seen, this solution first determines the conflict between empirical data and simulation / physical experimental data, then scores the credibility of conflicting simulation and physical experimental data, and finally deletes empirical data with lower credibility. This effectively filters out unreliable data information, ensuring the accuracy, consistency, and standardization of the target dataset. It provides high-quality and reliable data support for subsequent analysis and optimization of high-energy laser systems, avoiding the impact of data conflicts or unreliable data on the accuracy of subsequent work.

[0073] In one possible embodiment, generating a target dataset based on the experimental data and the target simulation data includes: fusing the experimental data and the target simulation data to obtain fused data; obtaining scene labels; performing deduplication processing on the fused data based on the scene labels to obtain reference fused data; and preprocessing the reference fused data to obtain the target dataset, wherein the preprocessing includes outlier removal and missing value imputation.

[0074] One approach is to use the Kalman filter algorithm to fuse the twin simulation data and the actual experimental data. Deduplication can include: using the standard deviation multiple of the 3σ criterion (default 3); assuming that all parameter data follow a normal distribution, using the 3σ criterion to remove outliers; if the Shapiro-Wilk normality test (P < 0.05) determines that the data does not follow a normal distribution, then automatically switching to the box plot method.

[0075] Based on the 3σ criterion, the sample mean μ of each parameter is first calculated. x With sample standard deviation σ x Using the Bessel correction formula, when the parameter value x i Satisfy |x i -μ x |>3σ x When an outlier is detected, it is identified and removed. The calculation formula is as follows: Let x be the sample mean of parameter x, and n be the total number of samples for parameter x. Let be the parameter value for the i-th sample.

[0076] Let x be the sample standard deviation of parameter x.

[0077] Missing value imputation can include linear interpolation, mean imputation, and median imputation.

[0078] Linear interpolation is used to fill missing data. First, the multi-source fusion standardized dataset is grouped by scene label, and within the same scene, it is sorted in ascending order by collection time / sample number. Then, the data values ​​and confidence score fields are traversed row by row. Explicit null values ​​and implicit invalid values ​​with confidence scores below 0.8 are marked as missing data and their sample numbers are recorded. Simultaneously, only missing points located in the middle of the data sequence with valid data in both preceding and following positions are selected as suitable for linear interpolation. If the missing data is at the beginning or end of the sequence or consists of consecutive missing values, other filling methods are switched according to the configuration. The filling value is derived by utilizing the linear relationship between the missing value and the adjacent valid data, ensuring the continuity of the data sequence. This method is suitable for scenarios where parameter values ​​change steadily with sample numbers. The specific calculation formula is as follows: k is the index of the missing sample. The filling result for the k-th missing value. For the first The parameter values ​​of each valid sample. The parameter value is for the (k+1)th valid sample.

[0079] After preprocessing, Z-score standardization can be used to eliminate dimensional differences between different parameters. Specifically, this involves converting the original parameter values ​​into dimensionless values ​​with a mean of 0 and a standard deviation of 1, facilitating multi-parameter comprehensive analysis and algorithmic calculations. The specific calculation formula is as follows: x is the original parameter value. Let x be the sample mean of parameter x. Let x be the sample standard deviation of parameter x. These are the standardized parameter values. The standardized data undergoes range validation to ensure that over 99% of the data falls within the [-3,3] interval. Data outside this interval is marked as suspicious and stored separately.

[0080] In one possible embodiment, determining a key parameter and its parameter influence from the plurality of parameter samples includes: obtaining the main effect value of each parameter from the plurality of parameter samples corresponding to each parameter, wherein the main effect value indicates the independent influence of each parameter; determining a reference parameter based on the main effect value; determining the interaction variance of each reference parameter pair from the plurality of parameter samples corresponding to the reference parameter, wherein the interaction variance indicates the contribution of the interaction between the reference parameter pairs to job performance, wherein the reference parameter pair includes any two reference parameters; and determining a key parameter from the reference parameters based on the interaction variance, wherein the parameter influence of the key parameter includes the main effect value and the interaction variance.

[0081] Before obtaining the main effect value of each parameter based on multiple parameter samples corresponding to each parameter, a standardization effect verification can be performed on all parameter samples to ensure that the mean of each parameter is in the range of [-0.1, 0.1], the standard deviation is in the range of [0.9, 1.1], and more than 99% of the parameter sample data fall within the range of [-3, 3]. If any verification criterion is not met, the parameter samples can be reprocessed to ensure that the parameter samples used to calculate the main effect value have uniform dimensions and good data quality, laying the foundation for the accuracy of subsequent main effect value calculation.

[0082] When determining reference parameters based on the main effect values, the parameters can be sorted in descending order of the absolute value of the main effect values. Parameters with larger absolute values ​​of the main effect values ​​and more significant independent impact on job performance can be selected as reference parameters. This can be done by setting a threshold for the main effect values ​​or by directly selecting a preset number of parameters that rank highly. This narrows the parameter range for subsequent interaction variance analysis, improves analysis efficiency, and retains parameters that have a significant independent impact on job performance to ensure the effectiveness of subsequent coupling effect analysis.

[0083] As can be seen, this scheme first quantifies the independent influence of a single parameter on operational performance by calculating the main effect value of each parameter and then selects reference parameters. Next, it calculates the interaction variance of the reference parameter pairs to quantify the contribution of the coupling effect between parameters. Finally, it combines the interaction variance to determine the key parameters from the reference parameters and uses both the main effect value and the interaction variance as the parameter influence. This approach can take into account both the independent influence of parameters and the coupling effect between parameters, accurately selecting the key parameters that play a core role in the operational performance of the high-energy laser system. At the same time, it obtains the comprehensive influence of the key parameters on operational performance in a complete and quantitative manner, making the determination of key parameters and their influence more scientific and accurate.

[0084] In one possible embodiment, obtaining the main effect value of each parameter based on multiple parameter samples corresponding to each parameter includes: performing the following operations for each parameter to obtain the main effect value of each parameter: dividing the value of the current parameter into a high value level and a low value level; obtaining the number of first parameter samples with values ​​at the high value level and the number of second parameter samples with values ​​at the low value level among the multiple current parameter samples; obtaining the mean of a first operational effect coefficient at the high value level and the mean of a second operational effect coefficient at the corresponding low value level for the current parameter, wherein the operational effect coefficient is used to indicate the effect of the laser on the target; and determining the main effect value of the current parameter based on the mean of the first operational effect coefficient, the mean of the second operational effect coefficient, the number of first parameter samples, and the number of second parameter samples.

[0085] This can be achieved by dividing each parameter into high and low value levels according to the configuration parameters, calculating the average performance coefficient of each parameter at both high and low levels, and then calculating the main effect value M. Ei This reflects the strength of the independent influence of the parameter. The parameter level division indicates the high or low level quantile; for example, the default 75th quantile is the high level X. i+ The 25th percentile is at the low level X. i- The formula for calculating the main effect value is as follows: Among them, M Ei For the i-th parameter X i Main effect value; n + n - For each parameter Xi + X i- The sample size at time; y is the operation effect coefficient, which is the result of comparing the energy density of the laser action point with the target material threshold after the laser completes one operation on the target, i.e., a single simulation test, a single physical test of a single set of parameters, or a virtual simulation. The value ranges from 0 to 1, and the closer it is to 1, the better the effect of the laser on the target in a single operation.

[0086] As can be seen, this scheme divides each parameter into high and low value levels, counts the number of parameter samples at each level, calculates the mean of the operation effect coefficient at each level, and then combines these data to determine the main effect value of the parameter. This can accurately reflect the degree of independent influence of a single parameter on the operation performance of the high-energy laser system through standardized quantitative calculation methods, making the acquisition process of the main effect value more standardized and accurate, and providing a reliable quantitative basis for the subsequent screening of key parameters.

[0087] In one possible embodiment, determining the interaction variance of each reference parameter pair based on multiple parameter samples corresponding to the reference parameter includes: obtaining the job performance parameter of each reference parameter sample corresponding to the reference parameter; determining the first job performance parameter mean of each reference parameter based on the job performance parameter of each reference parameter sample; obtaining the number of value levels of each reference parameter and the parameter sample corresponding to each value level; obtaining the second job performance parameter mean of each reference parameter at each value level based on the parameter sample corresponding to each value level; and determining the interaction variance of each reference parameter based on the first job performance parameter mean, the second job performance parameter mean, and the value level. The method involves: determining the single-parameter variance of each reference parameter; obtaining the level combinations of each reference parameter pair and the number of level combinations based on the value levels of each reference parameter; obtaining the mean of the third job performance parameter of each level combination and the mean of the fourth job performance parameter of each reference parameter pair based on the job performance parameters of each reference parameter sample; determining the parameter variance based on the mean of the third job performance parameter, the mean of the fourth job performance parameter, and the number of level combinations; and determining the interaction variance of each reference parameter pair based on the single-parameter variance of the reference parameters included in each reference parameter pair and the parameter variance of the reference parameter pair.

[0088] The calculation method for single-parameter variance, which reflects the performance fluctuation caused by different values ​​of a certain parameter, is as follows: Let be the variance of the i-th parameter. This refers to the number of value levels for this parameter. Value levels are used to quantify the impact of different values ​​of the same parameter on operational performance. The actual value range of the parameter is divided into discrete levels according to a preset rule; the 75th percentile is considered the high level. The 25th percentile is at a low level. ,Right now =2, Let be the number of samples at the j-th level of the i-th parameter. The average value of the performance parameters for the corresponding level; This represents the average performance parameter of all parameter samples for the i-th parameter. Performance parameters may include the performance coefficient, the duration of the operation, or the energy consumption. This can be the mean of the work effectiveness coefficient at the corresponding level, or the mean of the work duration, or the mean of the energy consumption. Therefore, the one-parameter variance of each parameter in terms of work effectiveness coefficient, work duration, or energy consumption can be obtained separately.

[0089] Interaction Variance (SSA) i×j The calculation formula is as follows: SSA i×j =SSA iAj -SSA i -SSA j In the formula, SSA i×j SSA represents the variance of the pairwise interaction between the i-th and j-th parameters. iAj The sum of squares of the two factors for the i-th and j-th parameters is calculated by grouping the groups according to the level combinations of parameters i and j, and then calculating the mean of the effect coefficients for each group. This can be obtained using the logic of single-parameter variance calculation. SSA i SSA j Let be the one-parameter variances of the i-th and j-th parameters, respectively.

[0090] As can be seen, this scheme first obtains the operational performance parameters of the reference parameter sample and calculates the relevant mean, then determines the variance of the single parameter and the variance of the parameters of the reference parameter pair step by step, and finally combines the two to calculate the variance of the interaction. Through a standardized and hierarchical quantitative calculation process, it can accurately decompose and quantify the actual contribution of the interaction between any two reference parameters to the operational performance of the high-energy laser system, making the analysis results of the coupling effect between parameters more objective and accurate, and providing reliable quantitative support for the accurate selection of key parameters in the future.

[0091] In one possible embodiment, determining the key parameter from the reference parameters based on the interaction variance includes: obtaining the job performance parameter of each reference parameter sample corresponding to the reference parameter and a first number of reference parameter samples; determining the mean of a fourth job performance parameter of all reference parameter samples based on the job performance parameter of each reference parameter sample; determining the total variance of the reference parameters based on the first number, the job performance parameter of each reference parameter sample, and the mean of the fourth job performance parameter; determining the individual parameter variance of each reference parameter; obtaining the error variance of all reference parameters; determining the total interaction variance of the reference parameters based on the total variance of the reference parameters, the individual parameter variances, and the error variances; and determining the key parameter from the reference parameters based on the interaction variance, the total variance of the reference parameters, and the total interaction variance.

[0092] Among them, if Significant, meaning the total interaction variance is greater than a preset value, allows for pairwise decomposition, with the SSA calculated for each parameter pair individually. i×j And determine a preset threshold based on the total variance of the reference parameter, for example, a preset threshold of SST × 5%, and then press SSA. i×j ≥SST×5% filters for key interaction items, ultimately yielding key parameters, with only the core SSAs being selected. i×j The corresponding interaction terms are included in the regression model, which ensures the accuracy of the analysis and avoids model redundancy.

[0093] The formula for calculating the total variance SST of the reference parameter is as follows: Where ȳ is the overall mean of the performance parameters, i.e., the mean of the fourth performance parameter; y i Let y be the job performance parameter for the i-th sample; i This is the labeling of the job performance coefficient of the i-th independent sample in the clean dataset formed after multi-source fusion and preprocessing, where i is the sequence index of the sample (i=1,2,3...n, n is the total number of samples). i It can be obtained from actual measurements in twin simulation experiments, and is obtained after multi-source data fusion and preprocessing, reflecting the true value of the actual operation effect of the sample; Total interaction variance The calculation formula is as follows: Where SST is the total variance, k is the total number of parameters involved in the analysis, and SSA is the total variance. i SSE represents the variance of all single parameters, and SSE represents the error variance.

[0094] When calculating the error variance, all samples can be divided according to the parameter level combination; the sum of squared deviations of all samples in each group from the mean of the group can be calculated; the sum of the squared deviations of all groups within the mean can be obtained as SSE.

[0095] As can be seen, this scheme first calculates the total variance of the reference parameter, the variance of each parameter, and the variance of the error, and then derives the total variance of the interaction. By combining the variance of the interaction, the total variance of the reference parameter, and the total variance of the interaction to screen key parameters, it can complete the scientific decomposition and quantitative comparison of variance based on the total variance. This allows the screening of key parameters to be based on a comprehensive variance analysis, accurately identifying reference parameters with significant coupling effects as key parameters, and improving the scientificity and accuracy of key parameter screening.

[0096] In one possible embodiment, the multi-objective optimization based on the key parameters, the parameter influence, the multi-objective optimization objective, and the engineering constraints includes: constructing a regression equation; obtaining the actual operational performance parameters of each key parameter sample corresponding to the key parameters; calculating the predicted operational performance parameters of each key parameter sample based on the regression equation; optimizing the regression equation based on the predicted operational performance parameters and the actual operational performance parameters to obtain a target regression equation, which is used to predict the operational performance parameters of the parameter samples; obtaining a single-objective fitness function based on the historical best operational performance parameters, the maximum allowable energy consumption of the project, and the longest allowable operation time; and performing multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the target regression equation to obtain the optimal combination of target parameters.

[0097] The formula for the regression equation is as follows: Where y is the performance parameter of the prediction task, a0 is a constant term, and a i b is the linear coefficient of the i-th parameter, reflecting the strength of the independent influence of the parameter. ij The interaction coefficient between the i-th and j-th parameters reflects the strength of the coupling effect; k is the total number of parameters; and ε is the random error term (satisfying a normal distribution ε~N(0,σ)). 2 )), X i Let a0 and a be the i-th input parameter. i b ij Let be the unknown coefficient to be solved. i The larger the absolute value of X, the stronger the parameter X. i The stronger the independent influence on the target indicator; i The positive or negative sign indicates the influence on the trend; a positive sign means that as the parameter increases, the indicator increases; a negative sign means that as the parameter increases, the indicator decreases. ij The larger the absolute value of X, the stronger the parameter X. i With X j The stronger the coupling and interaction, the greater the influence; ij The positive and negative values ​​indicate the interactive trend. A positive value means that both parameters increase at the same time, and the indicators have a positive superposition effect.

[0098] Model coefficient solution objective (minimize the sum of squared residuals): Where Q is the sum of squared residuals. Let be the actual operational performance parameters of the i-th sample. Here, n represents the total number of parameter samples, and the model predicts the job performance parameters.

[0099] For the work effectiveness coefficient, energy consumption, and work duration, the regression model for each indicator can be solved independently using the least squares method. The coefficients of the three models are independent of each other, and the core objective of solving them is to minimize the sum of squared residuals between the model predictions and the actual sample values. The solution process reuses the standardized process of the multi-factor influence sensitivity analysis module, with only the actual sample values ​​of the target indicators being different.

[0100] The optimization objective of this solution can be a single-objective or multi-objective requirement specified by the user. For example, a single objective is to maximize the work efficiency coefficient y; a multi-objective objective is to maximize y and minimize energy consumption E. The constraints include: hardware constraints: upper and lower limits of parameter values; performance constraints: upper limit of energy consumption E. max Maximum working time t max The sum of the weight coefficients for each objective is 1.

[0101] Optimize the objective and constraint analysis, clarify the core objective, and transform the constraints into mathematical expressions. Multi-objective optimization: The weighted summation method is used to transform multi-objectives into a single-objective fitness function, as shown in the following formula. f is the fitness value (the larger the value, the better the parameter combination); where: ① The historical best performance coefficient (derived from sample data statistics); ② The maximum energy consumption allowed by the project; ③ The maximum allowed duration for the task is w1, w2, and w3 are weighting coefficients; y is the task effectiveness coefficient.

[0102] As can be seen, this scheme first constructs and optimizes an accurate target regression equation based on actual and predicted operational performance parameters, then constructs a single-objective fitness function by combining historical best operational performance parameters and engineering limits, and finally conducts multi-objective optimization based on this function, engineering constraints, and the target regression equation. This allows multi-objective optimization to be based on accurate performance prediction, unifies multi-objective evaluation standards with a quantified fitness function, and strictly follows engineering constraints, significantly improving the scientific nature and accuracy of multi-objective optimization. It ensures that the obtained optimal combination of target parameters not only meets the requirements of actual engineering applications but also achieves the comprehensive optimization of multi-dimensional optimization objectives.

[0103] In one possible embodiment, the step of performing multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the objective regression equation to obtain the optimal parameter combination includes: obtaining multiple key parameter combinations based on the key parameters; obtaining the parameter contribution rate of each key parameter; encoding each key parameter combination into a chromosome based on the parameter contribution rate to obtain an initial population; and substituting the key parameter combination corresponding to each chromosome in the current population into the objective regression equation to obtain the reference job performance parameters output by the objective regression equation. In the initial iteration, the current population is the initial population, and the reference job performance parameters include job effectiveness coefficients, job duration, or energy. The process involves several steps: First, the operational effectiveness coefficient is used to indicate the effect of the laser on the target. Second, a reference fitness value is determined based on the single-target fitness function according to the reference operational performance parameters. Third, the next generation population is determined based on the reference fitness value. Fourth, the next generation population is updated using a single-point crossover strategy and a bit mutation strategy. Fifth, the next generation population is determined as the current population. The above steps are iteratively executed until a preset convergence condition is met. This preset convergence condition includes that the change in the reference fitness value of consecutive preset generations of populations in terms of the operational effectiveness coefficient, operational duration, and / or energy consumption is less than a second preset value. Sixth, the optimal combination of target parameters is determined based on the chromosomes in the last generation population and the engineering constraints.

[0104] The optimization algorithm parameters in this scheme can be: population size N (default 80), crossover probability pc (default 0.8), mutation probability p_m (default 0.05), convergence threshold (fitness change ≤ 1-4 over 20 consecutive generations), and maximum number of iterations (default 100 generations).

[0105] The specific optimization process includes: Population initialization: Key parameters are combined and encoded into binary chromosomes. The number of encoding bits is determined by combining the parameter contribution rate and the allowable engineering error: ① For key parameters with a contribution rate ≥ 40%, the number of encoding bits m satisfies the following conditions: ; ② 30%-40% of the parameters, engineering error ≤ 0.5%, default 16 bits; ③ For parameters less than 30%, the engineering error is ≤1%; users can customize and adjust the number of bits in the encoding according to their actual accuracy requirements.

[0106] The parameter contribution rate can be calculated using the following formula: Among them, SSA i Let S be the one-parameter variance of the i-th key parameter, and SST be the total variance.

[0107] The parameters corresponding to each chromosome are substituted into the multiple linear regression model of the multi-factor influence sensitivity analysis module. The target parameters are used as input parameters to solve the prediction model, obtaining the model prediction value for each type of target parameter. The task effectiveness coefficient, energy consumption, and task duration are calculated, and the fitness value is calculated based on the fitness function. A roulette wheel selection algorithm is used, allocating selection probabilities according to the proportion of fitness values, selecting individuals with high fitness to enter the next generation. A single-point crossover strategy is used, setting crossover points at random locations on chromosomes to exchange partial gene segments between two parent individuals, with a crossover probability of 0.8. A locus mutation strategy is used, randomly flipping some gene loci on chromosomes, with a mutation probability of 0.05, to avoid the algorithm getting trapped in local optima. If the fitness change is ≤1-4 for 20 consecutive generations, the optimization stops.

[0108] As can be seen, this scheme generates an initial population by encoding key parameter combinations according to their contribution rates. The population parameters are then substituted into the target regression equation to obtain reference job performance parameters and calculate reference fitness values. The population is iteratively updated through selection, single-point crossover, and position mutation strategies until the convergence condition is met. Finally, the optimal parameter combination is determined from the final population by combining engineering constraints. This approach allows multi-objective optimization to focus on key parameters and encode them differently according to their contribution rates. The iterative search of the genetic algorithm efficiently mines the optimal solution in the multi-dimensional parameter space. At the same time, the convergence condition controls the optimization accuracy, and the engineering constraints ensure feasibility. This significantly improves the efficiency and accuracy of multi-objective optimization, ensuring that the obtained optimal parameter combination has both optimal comprehensive performance and practical engineering feasibility.

[0109] In one possible embodiment, the method further includes: if the preset convergence condition is not met after iteratively executing the above steps for the Nth time, increasing the current population size and increasing the crossover probability of the single-point crossover strategy, and continuing to iterate for M times until the preset convergence condition is met, where N and M are positive numbers; if the preset convergence condition is not met after continuing to iterate for M times, using a simulated annealing algorithm to optimize and iteratively execute the above steps again based on the optimal parameter combination obtained in the Mth iteration.

[0110] In practice, if the population does not converge after 100 iterations, the population size is increased to 100, the crossover probability is adjusted to 0.85, and the iteration continues for another 50 generations. If the population still does not converge, local optimization and completion are enabled, and the simulated annealing algorithm is used to optimize the parameter space around the optimal individual, supplementing up to 30 generations.

[0111] As can be seen, this scheme can effectively avoid the genetic algorithm getting stuck in local optima by increasing the population size and crossover probability and continuing to iterate for M times when it fails to converge after N iterations. If it still fails to converge, it can further optimize the optimal parameter combination by combining simulated annealing algorithm. This can broaden the parameter search range and improve the global search capability, ensuring that the multi-objective optimization process can fully explore the optimal solution in the multi-dimensional parameter space, and further improve the accuracy and reliability of the optimal parameter combination of the target.

[0112] In one possible embodiment, the step of constructing a multi-factor optimization knowledge base based on the target optimal parameter combination includes: obtaining scene labels for the target optimal parameter combination; inputting the target optimal parameter combination into the digital twin based on the scene labels for multiple simulation verifications to obtain target performance data for the target optimal parameter combination; and constructing the multi-factor optimization knowledge base based on the target optimal parameter combination, the scene labels, and the target performance data.

[0113] The construction of the multi-factor optimization knowledge base includes the following: Knowledge base architecture setup: A data storage cluster is built based on the HBase distributed database, supporting simple queries with a response time of ≤1 second and complex relational queries with a response time of ≤3 seconds for databases of 100,000 records. A knowledge graph is built based on the Neo4j graph database, employing a synchronization strategy of real-time triggering and timed verification. After new parameter combinations, scene tags, and performance data are written to the HBase database, an event notification is sent via ZooKeeper to trigger the Neo4j knowledge graph to be updated synchronously, with a synchronization latency of ≤500ms. A full data verification is performed every morning at midnight, comparing the core relational data (scene ID-parameter combination-effect coefficient) between HBase and Neo4j. If inconsistencies are found, the HBase data is used as the standard and automatically corrected. If data conflicts occur during synchronization, query permissions for the conflicting data are temporarily locked and unlocked after synchronization is complete.

[0114] Initial knowledge storage: The optimal parameter combination, scenario tags, and performance data are written into the database according to the HBase table structure, establishing a relationship between data IDs and scenario IDs. Performance data, including job effectiveness coefficients, energy consumption, and job duration, are the core performance indicators for matching the optimal parameter combination. The association between the optimal parameter combination and the scenario is achieved by first defining a unique scenario tag and scenario feature parameters for each job scenario. During database entry, the optimal parameter combination, corresponding scenario tag, and performance data are uniformly written according to the HBase preset table structure. A basic association is established by creating a unique binding relationship between data IDs and scenario IDs. Then, four types of nodes—scenario tags, key parameters, optimal combinations, and job effectiveness—are created in the Neo4j knowledge graph. A knowledge link is constructed through directed relationships between scenario tags, key parameters, optimal combinations, and job effectiveness, allowing the optimal parameter combination to form a direct hierarchical association with the scenario tag, achieving precise mapping and traceability between specific scenarios and their corresponding optimal parameter combinations and performance data.

[0115] In one possible embodiment, after constructing a multi-factor optimization knowledge base based on the target optimal parameter combination, the method further includes: obtaining the current optimal parameter combination corresponding to the scene label; inputting the current optimal parameter combination into the digital twin based on the scene label for multiple simulation verifications to obtain the current performance data of the current optimal parameter combination; and updating the multi-factor optimization knowledge base based on the current optimal parameter combination, the scene label, and the current performance data.

[0116] The dynamic update mechanism can be triggered when the amount of new data reaches 10% of the total existing data, or when a new scene tag is detected.

[0117] The update process includes: calling the data preprocessing module to clean and standardize the new data; calling the multi-factor sensitivity analysis module to re-perform main effect analysis, variance decomposition, and interaction screening based on the updated dataset, and updating the parameter contribution rate and key parameter list; updating the regression model based on the new key parameter list; calling the runtime optimization design module to optimize for the new data / scenario and generate a new optimal parameter combination; writing the new parameter combination, scenario labels, and performance data into the HBase database and synchronously updating the nodes and relationships of the Neo4j knowledge graph; verifying the query accuracy of the updated knowledge base and generating an update log.

[0118] Please see Figure 2 The overall process of this application may include: The overall process of this solution is divided into two parallel main lines, which ultimately complete the closed loop through a trigger update judgment: The left main line starts with high-precision triple mapping and sequentially executes the steps of digital twin construction, data acquisition and fusion, data preprocessing, and dynamic update to realize the construction of a virtual model of the high-energy laser system, the integration and processing of multi-source data, and the real-time iteration of the model; The right main line starts with the input basic data and requirements and sequentially executes the steps of Kalman filter bias correction, combinatorial analysis, sensitivity analysis, intelligent optimization, knowledge base construction, and knowledge query to complete the scientific analysis of parameters, multi-objective optimization, and the accumulation and reuse of optimization knowledge; The outputs of the two main lines converge into the trigger update judgment link. If the update conditions are met, the corresponding step is returned for iterative optimization; if the update conditions are not met, the parameter optimization ends, forming a complete closed-loop process for multi-factor optimization of the high-energy laser system.

[0119] This application also provides a multi-factor optimization knowledge base construction system for high-energy laser systems. The system includes a Unity 3D digital twin construction module, a multi-source data acquisition and fusion module, a data preprocessing module, a multi-factor influence sensitivity analysis module, an operational state optimization design module, and a knowledge base construction and updating module.

[0120] The Unity 3D digital twin building module serves as the core simulation platform for the entire optimization system. Based on the Unity3D engine, it achieves high-precision geometric, physical, and behavioral mapping of the high-energy laser operation system, environment, and target, providing a highly reliable virtual experimental scenario for subsequent data acquisition, analysis, and optimization. Inputs include geometric modeling data, physical attribute parameter tables, and behavioral logic requirement files. Outputs include a Unity3D scene file containing the laser operation system model, target object model, environment scene model, and standardized API interface documentation. The standardized API interface documentation includes interface addresses, request methods, parameter formats, return examples, and error code explanations.

[0121] The Unity 3D Digital Twin building block can be used for standardized interface development, specifically including: developing RESTful API interfaces using HTTP / HTTPS protocols and supporting JSON format data transmission; input interfaces: receiving external parameter commands, such as laser power P=400kW, target velocity v=300m / s, and environmental visibility V=8km), with an interface response time ≤100ms; control interfaces: providing three types of control commands: simulation start / stop, scene switching, and parameter reset, supporting synchronous / asynchronous calls; and real-time push of simulation data with a data update frequency of 10Hz.

[0122] The multi-source data acquisition and fusion module is responsible for collecting three types of data: digital twin simulation, physical experiments, and empirical knowledge. It uses data fusion algorithms to eliminate data noise and systematic errors, generating a standardized dataset that provides high-quality input for subsequent data preprocessing. Input includes experimental data and target simulation data from the digital twin. Output includes a standardized multi-source fusion dataset: a JSON file containing four types of parameters: laser system, target, environment, and operational performance. The data reliability score is ≥0.8, and the deviation from the true value is ≤3%. Output may also include a data fusion report: a Word document containing statistics on data collection volume, fusion algorithm parameter settings, data deviation analysis, and records of outlier data removal.

[0123] The multi-source data acquisition and fusion module can also define a unified JSON data format, including fixed fields: data ID, parameter category, parameter name, parameter symbol, value, unit, scene label, acquisition time, data type, and credibility score. The fused data undergoes deduplication, based on a combination of data ID and scene label, to generate the final standardized dataset.

[0124] The data preprocessing module is used to remove outliers, impute missing values, and standardize units in the multi-source fusion standardized dataset, eliminating data noise and dimensional differences, and outputting a clean dataset to provide high-quality data support for subsequent multi-factor sensitivity analysis. Inputs include the multi-source fusion standardized dataset: a JSON file output by the multi-source data acquisition and fusion module, containing the original values ​​of all parameters, scene labels, data types, etc. Preprocessing configuration parameters can also be included as input. Outputs include a clean dataset: a JSON file containing original parameter values, preprocessed parameter values, data status, standardized values, etc., with data integrity ≥98.5%, and no significant outlier residue after multiple outlier removal and verification processes. Outputs can also include a preprocessing report: a Word document containing data cleaning statistics, the number of outliers, the number of missing values, standardized parameters, and data quality assessment.

[0125] The multi-factor sensitivity analysis module is the core analysis unit. It quantifies the contribution of individual parameters' independent influence and interactions between parameters to operational performance based on cleanroom datasets, identifies key parameters, and builds predictive models, providing clear direction and mathematical support for intelligent optimization. Inputs can include a JSON file output from the data preprocessing module, containing standardized laser system parameters, target parameters, environmental parameters, and corresponding operational performance indicators. Inputs can also include analysis configuration parameters, including target performance indicators (user-specified indicators to be analyzed); parameter level divisions (high / low quantiles); and regression model requirements (goodness of fit R-squared). 2Minimum threshold. Output may include a sensitivity analysis results set, including a JSON file containing the main effect values, contribution rates, key parameter lists, multiple linear regression model coefficients, and model fit R-squared. 2 Output may also include visualization charts, including PNG format, containing Pareto plots, main effect plots, and contour plots, with a resolution ≥1920×1080. Output may also include analysis reports, including Word format, containing the analysis process, the basis for identifying key parameters, regression model formulas, and interpretation of results.

[0126] The operational optimization design module is used to identify key parameters based on sensitivity analysis. It employs a genetic algorithm to search for the optimal parameter combination that satisfies the operational objectives and constraints in a multi-dimensional parameter space, ensuring the optimization results possess both engineering feasibility and optimal performance. Inputs include key parameter data, such as the list of key parameters output by the multi-factor influence sensitivity analysis module (including parameter names, value ranges, and contribution rates), and a multiple linear regression model. Inputs may also include optimization objectives and constraints. Outputs include the optimal parameter combination, presented in an Excel file containing key parameter names, optimal values, the basis for these values, and validation results including the mean and standard deviation of six simulations. Outputs may also include optimization process data, presented in a JSON file containing the optimal fitness value, average fitness value, and iteration curve data for each generation of the population. Outputs may also include an optimization report, presented in Word format, containing the optimization objective, constraints, algorithm parameters, optimization process, result validation conclusions, and engineering feasibility analysis.

[0127] The knowledge base construction and update module is used to realize the structured storage, related query and dynamic update of scenario-parameter-effect optimization knowledge, to build a reusable and iterative knowledge base, and to support rapid response to parameter optimization needs in different scenarios.

[0128] The inputs for initializing the knowledge base include: Optimal parameter combination data: an Excel file output by the running state optimization design module, containing parameter names, optimal values, and verification results; Scene tag data: user-defined scene category tags, scene feature parameters, target materials, environmental conditions, and job requirements; Performance data: job effect coefficient y, energy consumption E, and job duration t after optimization and verification; Knowledge base configuration parameters: HBase database connection parameters (IP, port, username, password), and Neo4j knowledge graph node / relationship definition rules.

[0129] The dynamically updated inputs are: New datasets: New twin simulation data, entity test data, and experiential knowledge data, with the same format as the input to the multi-source data acquisition and fusion module; New scene labels: User-defined new scene classification labels and feature parameters.

[0130] The initial output of the knowledge base includes: an HBase distributed database storing 100,000 simulation / experiment data points; a Neo4j knowledge graph containing four types of nodes—scenes, parameters, combinations, and effects—and their relationships; a web query interface supporting fuzzy and precise queries based on scene tags, and a RESTful API query interface, including interface documentation; and a knowledge base initialization report in Word format, including the database structure design, data volume statistics, and query performance test results.

[0131] The dynamic update output includes: the updated knowledge base containing new data and knowledge entries corresponding to new scenario tags, with knowledge graph nodes / relationships updated synchronously; knowledge base update log: Excel format, including update trigger conditions, amount of new data, update time, and verification results; update report: Word format, including update process, knowledge iteration content, and query performance change analysis.

[0132] As can be seen from the above, this solution addresses the shortcomings of existing technologies that lack data reliability and coverage by providing simulation and data support from a singular to a comprehensive and high-precision approach. Existing technologies only optimize single environmental parameters and do not involve full-scene simulation of the system, environment, and target; they rely on limited physical experimental data, and the simulation focuses only on a single physical field, such as laser energy transmission, lacking geometric and behavioral mapping, resulting in data deviations exceeding 15%. The technical advantages of this application are: constructing a digital twin with triple mapping of geometry, physics, and behavior using Unity3D; simultaneously integrating three types of data—twin simulation, physical experiments, and empirical knowledge—and employing a Kalman filter algorithm to reduce errors, resulting in a data deviation of ≤3% from the true value after fusion. The data covers all dimensions of parameters, including laser system, target, environment, and performance.

[0133] This solution addresses the shortcomings of existing technologies that fail to accurately identify key parameters, moving from single-dimensional to coupled quantification in multi-factor analysis. These technologies often lack multi-parameter coupled analysis, focusing only on optimizing single environmental parameters and failing to quantify parameter interactions. Existing technologies typically employ single statistical analysis methods, such as analysis of variance, resulting in key parameter identification accuracy below 70% and lacking visualization support. This application innovatively employs a combined analysis system of main effect analysis, analysis of variance, and multiple linear regression. This system quantifies the independent main effect value of individual parameters while accurately capturing the contribution and variance of interaction effects between parameters to performance. Visual outputs such as Pareto charts and main effect diagrams assist engineers in quickly identifying core factors. The key parameter identification accuracy is improved to over 90%, and the regression model's goodness of fit R0 is significantly increased. 2 With a value ≥0.85 and a prediction error ≤5%, it far exceeds the analytical depth and accuracy of existing technologies.

[0134] Furthermore, this solution addresses the inefficiencies and singular objectives of optimization by moving from experience-driven to data-algorithm-based approaches. Existing technologies often employ manual debugging and trial-and-error methods, leading to aimless optimization and susceptibility to local optima. While some solutions introduce simple optimization algorithms, they lack integration with multi-factor analysis results, failing to address multiple objectives such as performance and energy consumption balance, resulting in optimization cycles of 7-15 days per scenario. The advantages of this application lie in its deep integration of the optimization process and sensitivity analysis results. It allocates genetic algorithm coding precision based on parameter contribution rates, with higher contribution rate parameters achieving higher coding precision. This focuses on key parameters, narrowing the search range and avoiding exhaustive search traps. It supports maximizing single-objective job performance and optimizing multi-objective performance and energy consumption balance. A fitness function is designed using a weighted summation method, taking into account engineering constraints. The number of optimization iterations is ≤180 generations, with the iteration rhythm dynamically adjusted based on the convergence threshold to ensure the algorithm approaches the global optimum. The optimization cycle for a single scenario is shortened to 1 day. The optimization results are verified by a digital twin, with a deviation of ≤5%, demonstrating stronger engineering feasibility.

[0135] From Scattered Storage to Structured Dynamic Accumulation: Solving the Problems of Lagging Knowledge Reuse and Adaptation Existing Technologies' Shortcomings: Existing technologies lack structured knowledge bases, storing optimization knowledge in unstructured forms such as experience logs and experiment reports. This results in broken relationships between scenarios, parameters, and effects, requiring repeated experiments and analyses for the same scenarios, leading to a knowledge reuse rate of less than 30%. Furthermore, dynamic iteration of knowledge cannot be achieved when new scenarios or data are added, necessitating the rebuilding of analysis and optimization models, resulting in extremely low adaptation efficiency. This application's technical advantages: Utilizing a ZooKeeper, HBase distributed database, and Neo4j knowledge graph architecture, it establishes clear relationships between scenario tags, key parameters, optimal combinations, and job effects, supporting second-level query response times of ≤1 second for 100,000-level data. A data-driven dynamic update mechanism is designed, automatically triggering an iterative update process when new data reaches 10% of the existing total or when new scenario tags are added. Adaptation time for new scenarios is ≤2 hours, establishing a full-link relationship between scenarios, parameters, and effects, increasing the knowledge reuse rate to over 62%, and significantly reducing the cost of repeated experiments in the same or similar scenarios.

[0136] The following describes a multi-factor optimization knowledge base construction device for a high-energy laser system provided in this application. The multi-factor optimization knowledge base construction device for a high-energy laser system described below corresponds to the multi-factor optimization knowledge base construction method for a high-energy laser system described above.

[0137] Please see Figure 3The multi-factor optimization knowledge base construction device 300 for high-energy laser systems includes: an acquisition unit 301 for acquiring a target dataset, wherein the target dataset includes multiple parameters associated with the high-energy laser system, and each parameter corresponds to multiple parameter samples; a first determination unit 302 for determining key parameters and their parameter influence degrees from the multiple parameters based on the multiple parameter samples, wherein the parameter influence degree indicates the degree of contribution of the independent influence degree of each key parameter and the interaction between key parameters to the operational performance; a second determination unit 303 for determining multi-objective optimization objectives and engineering constraints; a multi-objective optimization unit 304 for performing multi-objective optimization based on the key parameters, the parameter influence degrees, the multi-objective optimization objectives, and the engineering constraints to obtain the optimal parameter combination; and a construction unit 305 for constructing a multi-factor optimization knowledge base based on the optimal parameter combination.

[0138] In one possible embodiment, in determining the key parameter and the parameter influence of the key parameter from the plurality of parameters based on the plurality of parameter samples, the first determining unit 302 is specifically configured to: obtain the main effect value of each parameter based on the plurality of parameter samples corresponding to each parameter, the main effect value being used to indicate the independent influence of each parameter; determine a reference parameter based on the main effect value; determine the interaction variance of each reference parameter pair based on the plurality of parameter samples corresponding to the reference parameter, the interaction variance being used to indicate the contribution of the interaction between the reference parameter pairs to job performance, the reference parameter pair including any two reference parameters; and determine the key parameter from the reference parameters based on the interaction variance, the parameter influence of the key parameter including the main effect value and the interaction variance.

[0139] In one possible embodiment, in obtaining the main effect value of each parameter based on multiple parameter samples corresponding to each parameter, the first determining unit 302 is specifically configured to: perform the following operations for each parameter to obtain the main effect value of each parameter: divide the value of the current parameter into a high value level and a low value level; obtain the number of first parameter samples with values ​​at the high value level and the number of second parameter samples with values ​​at the low value level among the multiple current parameter samples; obtain the mean of the first operational effect coefficient at the high value level corresponding to the current parameter and the mean of the second operational effect coefficient at the corresponding low value level, wherein the operational effect coefficient is used to indicate the effect of the laser on the target; and determine the main effect value of the current parameter based on the mean of the first operational effect coefficient, the mean of the second operational effect coefficient, the number of the first parameter samples, and the number of the second parameter samples.

[0140] In one possible embodiment, in determining the interaction variance of each reference parameter pair based on multiple parameter samples corresponding to the reference parameter, the first determining unit 302 is specifically configured to: obtain the job performance parameter of each reference parameter sample corresponding to the reference parameter; determine the first job performance parameter mean of each reference parameter based on the job performance parameter of each reference parameter sample; obtain the number of value levels of each reference parameter and the parameter sample corresponding to each value level; obtain the second job performance parameter mean of each reference parameter at each value level based on the parameter sample corresponding to each value level; and determine the interaction variance of each reference parameter based on the first job performance parameter mean and the second job performance parameter mean. The single-parameter variance of each reference parameter is determined by the value and the number of value levels; the level combinations of each reference parameter pair are obtained based on the value levels of each reference parameter, as well as the number of level combinations of the reference parameter pairs; the mean of the third job performance parameter of each level combination and the mean of the fourth job performance parameter of each reference parameter pair are obtained based on the job performance parameters of each reference parameter sample; the parameter variance is determined based on the mean of the third job performance parameter, the mean of the fourth job performance parameter, and the number of level combinations; the interaction variance of each reference parameter pair is determined based on the single-parameter variance of the reference parameters included in each reference parameter pair and the parameter variance of the reference parameter pair.

[0141] In one possible embodiment, in determining the key parameter from the reference parameters based on the interaction variance, the first determining unit 302 is specifically configured to: obtain the job performance parameter of each reference parameter sample corresponding to the reference parameter and a first number of the reference parameter samples; determine the mean of a fourth job performance parameter of all reference parameter samples based on the job performance parameter of each reference parameter sample; determine the total variance of the reference parameters based on the first number, the job performance parameter of each reference parameter sample, and the mean of the fourth job performance parameter; determine the individual parameter variance of each reference parameter; obtain the error variance of all reference parameters; determine the total interaction variance of the reference parameters based on the total variance of the reference parameters, the individual parameter variance, and the error variance; and determine the key parameter from the reference parameters based on the interaction variance, the total variance of the reference parameters, and the total interaction variance.

[0142] In one possible embodiment, regarding the multi-objective optimization based on the key parameters, the parameter influence, the multi-objective optimization objective, and the engineering constraints, the multi-objective optimization unit 304 is specifically configured to: construct a regression equation; obtain the actual operational performance parameters of each key parameter sample corresponding to the key parameters; calculate the predicted operational performance parameters of each key parameter sample based on the regression equation; optimize the regression equation based on the predicted operational performance parameters and the actual operational performance parameters to obtain a target regression equation, wherein the target regression equation is used to predict the operational performance parameters of the parameter samples; obtain a single-objective fitness function based on the historical best operational performance parameters, the maximum allowable energy consumption of the project, and the longest allowable operational time; and perform multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the target regression equation to obtain the optimal combination of target parameters.

[0143] In one possible embodiment, regarding the multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the objective regression equation to obtain the optimal parameter combination, the multi-objective optimization unit 304 is specifically configured to: obtain multiple key parameter combinations based on the key parameters; obtain the parameter contribution rate of each key parameter; encode each key parameter combination into a chromosome based on the parameter contribution rate to obtain an initial population; and input the key parameter combination corresponding to each chromosome in the current population into the objective regression equation to obtain reference job performance parameters output by the objective regression equation, wherein the current population is the initial population during the initial iteration, and the reference job performance parameters include job performance parameters. The process involves considering factors such as the number of targets, operation time, or energy consumption, with the operation effect coefficient indicating the laser's effect on the target. A reference fitness value is determined based on the single-target fitness function according to the reference operation performance parameters. The next generation population is determined based on the reference fitness value. The next generation population is updated using a single-point crossover strategy and a bit mutation strategy. The next generation population is identified as the current population. The above steps are iteratively executed until a preset convergence condition is met. This preset convergence condition includes that the change in the reference fitness value of consecutive preset generations of populations in terms of the operation effect coefficient, operation time, and / or energy consumption is less than a second preset value. The optimal combination of target parameters is determined based on the chromosomes in the last generation population and the engineering constraints.

[0144] In one possible embodiment, the multi-objective optimization unit 304 is further configured to: increase the current population size and increase the crossover probability of the single-point crossover strategy after iterating through the above steps for the Nth time without satisfying the preset convergence condition, and continue iterating for M times until the preset convergence condition is satisfied, where N and M are positive numbers; if the preset convergence condition is not satisfied after iterating through M times, use the simulated annealing algorithm to optimize and iterate through the above steps again based on the optimal parameter combination obtained in the Mth iteration.

[0145] In one possible embodiment, in terms of acquiring the target dataset, the acquisition unit 301 is specifically configured to: construct a digital twin of a high-energy laser system; acquire experimental data and target simulation data of the digital twin; and generate a target dataset based on the experimental data and the target simulation data.

[0146] In one possible embodiment, in the step of constructing a multi-factor optimization knowledge base based on the target optimal parameter combination, the construction unit 305 is specifically used to: obtain scene labels for the target optimal parameter combination; input the target optimal parameter combination into the digital twin based on the scene labels for multiple simulation verifications to obtain target performance data of the target optimal parameter combination; and construct the multi-factor optimization knowledge base based on the target optimal parameter combination, the scene labels, and the target performance data.

[0147] In one possible embodiment, the multi-factor optimization knowledge base construction device 300 for a high-energy laser system further includes an update unit. After constructing the multi-factor optimization knowledge base based on the target optimal parameter combination, the update unit is specifically used to: obtain the current optimal parameter combination corresponding to the scene label; input the current optimal parameter combination into the digital twin based on the scene label for multiple simulation verifications to obtain the current performance data of the current optimal parameter combination; and update the multi-factor optimization knowledge base based on the current optimal parameter combination, the scene label, and the current performance data.

[0148] In one possible embodiment, in constructing the digital twin of the high-energy laser system, the acquisition unit 301 is specifically configured to: acquire geometric modeling data, a physical attribute parameter table, and a behavioral logic requirement file; construct a geometric twin model based on the geometric modeling data; configure physical twin characteristics based on the physical attribute parameter table; perform twin control based on the behavioral logic requirement file to obtain twin control results; and construct the digital twin based on the geometric twin model, the physical twin characteristics, and the twin control results.

[0149] In one possible embodiment, where the geometric modeling data includes 3D data of the laser operation system, 3D data of the target object, and component design drawings, the acquisition unit 301 is specifically used for: generating an initial model of the laser operation system based on the 3D data of the laser operation system; simplifying the initial model of the laser operation system into a mesh to obtain a simplified model of the laser operation system; calibrating and simulating the simplified model of the laser operation system based on the component design drawings to obtain a geometric twin model of the laser operation system; generating an initial model of the target object based on the 3D data of the target object; configuring the size parameters of the target object; generating a geometric twin model of the target object based on the size parameters and the initial model of the target object; and constructing a geometric twin model based on the geometric twin model of the laser operation system and the geometric twin model of the target object.

[0150] In one possible embodiment, where the experimental data includes physical experimental data and empirical data, and regarding the acquisition of experimental data and the target simulation data of the digital twin, the acquisition unit 301 is specifically used to: acquire real-time twin simulation data, output signals of experimental sensors, and experimental logs; parse the JSON fields of the real-time twin simulation data to obtain the target simulation data; convert the output signals into target digital signals to obtain physical experimental data; and acquire key-value pairs based on the experimental logs to obtain the empirical data.

[0151] In one possible embodiment, before generating the target dataset based on the experimental data and the target simulation data, the acquisition unit 301 is specifically configured to: determine whether there is a conflict between the empirical data and the target simulation data or the entity experimental data; if there is a conflict, determine the credibility score of the target simulation data or the entity experimental data; and delete the empirical data if the credibility score is higher than a preset score.

[0152] In one possible embodiment, in generating the target dataset based on the experimental data and the target simulation data, the acquisition unit 301 is specifically configured to: fuse the experimental data and the target simulation data to obtain fused data; acquire scene labels; perform deduplication processing on the fused data based on the scene labels to obtain reference fused data; and preprocess the reference fused data to obtain the target dataset, wherein the preprocessing includes outlier removal processing and missing value imputation processing.

[0153] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. For example... Figure 4As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for constructing a multi-factor optimization knowledge base for a high-energy laser system, including: acquiring a target dataset, the target dataset including multiple parameters associated with the high-energy laser system, each parameter corresponding to multiple parameter samples; determining key parameters and their parameter influence degrees from the multiple parameters based on the multiple parameter samples, the parameter influence degree indicating the degree of contribution of the independent influence degree of each key parameter and the interaction between key parameters to the operational performance; determining multi-objective optimization objectives and engineering constraints; performing multi-objective optimization based on the key parameters, the parameter influence degrees, the multi-objective optimization objectives, and the engineering constraints to obtain the optimal parameter combination; and constructing a multi-factor optimization knowledge base based on the optimal parameter combination.

[0154] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] On the other hand, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the multi-factor optimization knowledge base construction method for high-energy laser systems provided by the methods described above. The method includes: acquiring a target dataset, the target dataset including multiple parameters associated with a high-energy laser system, each parameter corresponding to multiple parameter samples; determining key parameters and their parameter influence degrees from the multiple parameters based on the multiple parameter samples, the parameter influence degree indicating the degree of independent influence of each key parameter and the contribution of the interaction between key parameters to operational performance; determining multi-objective optimization objectives and engineering constraints; performing multi-objective optimization based on the key parameters, the parameter influence degrees, the multi-objective optimization objectives, and the engineering constraints to obtain the optimal parameter combination; and constructing a multi-factor optimization knowledge base based on the optimal parameter combination.

[0156] In another aspect, this application also provides a computer program product, including a computer program that, when executed by a processor, implements a method for constructing a multi-factor optimization knowledge base for any of the high-energy laser systems described above, comprising: acquiring a target dataset, the target dataset including multiple parameters associated with the high-energy laser system, each parameter corresponding to multiple parameter samples; determining key parameters and the parameter influence degree of the key parameters from the multiple parameters based on the multiple parameter samples, the parameter influence degree indicating the degree of contribution of the independent influence degree of each key parameter and the interaction between key parameters to the operational performance; determining multi-objective optimization objectives and engineering constraints; performing multi-objective optimization based on the key parameters, the parameter influence degree, the multi-objective optimization objectives, and the engineering constraints to obtain the optimal parameter combination for the objective; and constructing a multi-factor optimization knowledge base based on the optimal parameter combination for the objective.

[0157] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0158] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for constructing a multi-factor optimization knowledge base for a high-energy laser system, characterized in that, include: Obtain the target dataset, which includes multiple parameters associated with the high-energy laser system, with each parameter corresponding to multiple parameter samples; Based on the multiple parameter samples, key parameters and their parameter influence are determined from the multiple parameters. The parameter influence indicates the degree of contribution of the independent influence of each key parameter and the interaction between key parameters to the job performance. Determine the multi-objective optimization objectives and engineering constraints; Based on the key parameters, the parameter influence, the multi-objective optimization objective, and the engineering constraints, multi-objective optimization is performed to obtain the optimal combination of objective parameters. A multi-factor optimization knowledge base is constructed based on the optimal combination of parameters for the target.

2. The method according to claim 1, characterized in that, The step of determining key parameters and the parameter influence degree of the key parameters from the multiple parameter samples includes: The main effect value of each parameter is obtained from multiple parameter samples corresponding to each parameter, and the main effect value is used to indicate the independent influence of each parameter; The reference parameters are determined based on the main effect values; The interaction variance of each reference parameter pair is determined based on multiple parameter samples corresponding to the reference parameters. The interaction variance is used to indicate the degree of contribution of the interaction between the reference parameter pairs to the job performance. The reference parameter pair includes any two reference parameters. Key parameters are determined from the reference parameters based on the interaction variance, wherein the parameter influence of the key parameters includes the main effect value and the interaction variance.

3. The method according to claim 2, characterized in that, The step of obtaining the main effect value of each parameter based on multiple parameter samples corresponding to each parameter includes: Perform the following operations for each parameter to obtain the main effect value of each parameter: Divide the current parameter value into high value level and low value level; Obtain the number of first parameter samples with values ​​at the higher level and the number of second parameter samples with values ​​at the lower level from the plurality of current parameter samples; The average value of the first operational effect coefficient at the high value level corresponding to the current parameter is obtained, and the average value of the second operational effect coefficient at the low value level is obtained. The operational effect coefficient is used to indicate the effect of the laser on the target. The main effect value of the current parameter is determined based on the mean of the first task effect coefficient, the mean of the second task effect coefficient, the number of samples for the first parameter, and the number of samples for the second parameter.

4. The method according to claim 2, characterized in that, The step of determining the interaction variance of each reference parameter pair based on multiple parameter samples corresponding to the reference parameters includes: Obtain the job performance parameters for each reference parameter sample corresponding to the reference parameters; The mean value of the first operational performance parameter for each reference parameter is determined based on the operational performance parameters of each reference parameter sample. Obtain the number of value levels for each reference parameter and the parameter sample corresponding to each value level; Based on the parameter samples corresponding to each value level, obtain the average value of the second operational performance parameter of each reference parameter at each value level; The single-parameter variance of each reference parameter is determined based on the mean of the first job performance parameter, the mean of the second job performance parameter, and the number of value levels. Based on the value level of each reference parameter, obtain the level combination of each reference parameter pair, and the number of level combinations of each reference parameter pair; The mean of the third job performance parameter for each level combination and the mean of the fourth job performance parameter for each reference parameter pair are obtained based on the job performance parameters of each reference parameter sample. The variance of the parameters is determined based on the mean of the third job performance parameter, the mean of the fourth job performance parameter, and the number of level combinations. The interaction variance of each reference parameter pair is determined based on the single-parameter variance of the reference parameter included in each reference parameter pair and the parameter variance of the reference parameter pair.

5. The method according to claim 2, characterized in that, The step of determining the key parameters from the reference parameters based on the interaction variance includes: Obtain the job performance parameters for each reference parameter sample corresponding to the reference parameters and the first number of the reference parameter samples; The mean of the fourth operational performance parameter of all reference parameter samples is determined based on the operational performance parameters of each reference parameter sample. The total variance of the reference parameters is determined based on the first quantity, the job performance parameters of each reference parameter sample, and the mean of the fourth job performance parameter. Determine the single-parameter variance of each of the reference parameters; Obtain the error variance of all the reference parameters; The interaction total variance of the reference parameter is determined based on the total variance of the reference parameter, the variance of the individual parameter, and the variance of the error. Key parameters are determined from the reference parameters based on the interaction variance, the total variance of the reference parameters, and the total interaction variance.

6. The method according to claim 1, characterized in that, The multi-objective optimization based on the key parameters, the parameter influence, the multi-objective optimization objective, and the engineering constraints includes: Construct a regression equation; Obtain the actual operational performance parameters for each key parameter sample corresponding to the key parameters; The predicted job performance parameters for each key parameter sample are calculated based on the regression equation. The regression equation is optimized based on the predicted job performance parameters and the actual job performance parameters to obtain the target regression equation, which is used to predict the job performance parameters of the parameter sample. The single-objective fitness function is obtained based on the historical best performance parameters, the maximum energy consumption allowed by the project, and the longest allowed operation time. Based on the single-objective fitness function, the engineering constraints, and the objective regression equation, multi-objective optimization is performed to obtain the optimal combination of objective parameters.

7. The method according to claim 6, characterized in that, The step of performing multi-objective optimization based on the single-objective fitness function, the engineering constraints, and the objective regression equation to obtain the optimal combination of objective parameters includes: Based on the aforementioned key parameters, obtain multiple combinations of key parameters; Obtain the parameter contribution rate for each key parameter; Each key parameter is combined and encoded into a chromosome based on the parameter contribution rate to obtain the initial population; The key parameters corresponding to each chromosome in the current population are combined and substituted into the target regression equation to obtain the reference operation performance parameters output by the target regression equation. In the initial iteration, the current population is the initial population. The reference operation performance parameters include the operation effect coefficient, operation time or energy consumption. The operation effect coefficient is used to indicate the effect of the laser on the target. The reference fitness value is determined based on the single-objective fitness function according to the reference job performance parameters; The next generation population is determined based on the reference fitness value; The next generation population is updated using a single-point crossover strategy and a positional mutation strategy. The next generation of the population is determined as the current population, and the above steps are iteratively executed until a preset convergence condition is met. The preset convergence condition includes that the change in the reference fitness value of the population in terms of the operation effect coefficient, the operation duration and / or the energy consumption of consecutive preset generations is less than a second preset value. The optimal combination of parameters for the objective is determined based on the chromosomes in the last generation population and the engineering constraints.

8. The method according to claim 7, characterized in that, The method further includes: If the preset convergence condition is not met after the Nth iteration of the above steps, the current population size is increased and the crossover probability of the single-point crossover strategy is increased. The iteration continues for M times until the preset convergence condition is met, where N and M are positive numbers. If the preset convergence condition is not met after M iterations, the simulated annealing algorithm is used to optimize and iterate the above steps again based on the optimal parameter combination obtained in the Mth iteration.

9. The method according to claim 1, characterized in that, The acquisition of the target dataset includes: Constructing a digital twin of a high-energy laser system; Acquire experimental data and target simulation data of the digital twin; A target dataset is generated based on the experimental data and the target simulation data.

10. The method according to claim 9, characterized in that, The construction of a multi-factor optimization knowledge base based on the target optimal parameter combination includes: Obtain the scene label of the optimal combination of parameters for the target; Based on the scene labels, the optimal combination of target parameters is input into the digital twin for multiple simulation verifications to obtain the target performance data of the optimal combination of target parameters; The multi-factor optimization knowledge base is constructed based on the target optimal parameter combination, the scene label, and the target performance data.

11. The method according to claim 10, characterized in that, After constructing a multi-factor optimization knowledge base based on the target optimal parameter combination, the method further includes: Obtain the current optimal parameter combination corresponding to the scene label; Based on the scene label, the current optimal parameter combination is input into the digital twin for multiple simulation verifications to obtain the current performance data of the current optimal parameter combination; The multi-factor optimization knowledge base is updated based on the current optimal parameter combination, the scene label, and the current performance data.

12. The method according to claim 9, characterized in that, The digital twin used to construct the high-energy laser system includes: Obtain geometric modeling data, physical property parameter tables, and behavioral logic requirement files; Construct a geometric twin model based on the geometric modeling data; Configure physical twin characteristics according to the physical property parameter table; Perform twin control based on the aforementioned behavioral logic requirement document to obtain the twin control result; The digital twin is constructed based on the geometric twin model, the physical twin characteristics, and the twin control results.

13. The method according to claim 12, characterized in that, The geometric modeling data includes 3D data of the laser operation system, 3D data of the target object, and component design drawings. The step of constructing a geometric twin model based on the geometric modeling data includes: An initial model of the laser operation system is generated based on the three-dimensional data of the laser operation system. The initial model of the laser operation system is simplified by meshing to obtain a simplified model of the laser operation system; The simplified model of the laser operation system was calibrated and simulated based on the component design drawings to obtain a geometric twin model of the laser operation system. Generate an initial model of the target object based on the three-dimensional data of the target object; Configure the size parameters of the target object; Generate a geometric twin model of the target object based on the size parameters and the initial model of the target object; A geometric twin model is constructed based on the geometric twin model of the laser operation system and the geometric twin model of the target object.

14. The method according to claim 9, characterized in that, The experimental data includes physical experimental data and empirical data. The acquisition of the experimental data and the target simulation data of the digital twin includes: Acquire real-time data from the twin simulation, output signals from experimental sensors, and experimental logs; The target simulation data is obtained by parsing the JSON fields of the twin simulation real-time data; The output signal is converted into a target digital signal to obtain physical experimental data; The key-value pairs are obtained from the experimental logs to obtain the empirical data.

15. The method according to claim 14, characterized in that, Before generating the target dataset based on the experimental data and the target simulation data, the method further includes: Determine whether there is any conflict between the empirical data and the target simulation data or the physical experiment data; If a conflict exists, a credibility score is determined for the target simulation data or the entity experimental data. If the credibility score is higher than the preset score, the empirical data is deleted.

16. The method according to claim 9, characterized in that, A target dataset is generated based on the experimental data and the target simulation data, including: The experimental data and the target simulation data are fused to obtain fused data; Get scene tags; The fused data is deduplicated based on the scene labels to obtain reference fused data; The reference fusion data is preprocessed to obtain the target dataset. The preprocessing includes outlier removal and missing value imputation.

17. A device for constructing a multi-factor optimization knowledge base for a high-energy laser system, characterized in that, include: An acquisition unit is used to acquire a target dataset, which includes multiple parameters associated with a high-energy laser system, and each parameter corresponds to multiple parameter samples; The first determining unit is configured to determine key parameters and the parameter influence degree of the key parameters from the multiple parameter samples, wherein the parameter influence degree indicates the degree of contribution of the independent influence degree of each key parameter and the interaction between key parameters to the job performance. The second determining unit is used to determine the multi-objective optimization objective and engineering constraints. A multi-objective optimization unit is used to perform multi-objective optimization based on the key parameters, the parameter influence degree, the multi-objective optimization objectives and the engineering constraints, to obtain the optimal combination of objective parameters. The construction unit is used to construct a multi-factor optimization knowledge base based on the optimal combination of the target parameters.

Images